Getting Started

In 10 minutes, you'll create your first Workbook, import data, talk to the AI agent, and understand how Views work. You'll see the SQL, but you won't write any.

---

Part 1: Create Your Workbook (2 min)

What's a Workbook?

A Workbook is your analytical workspace for a project. It holds:

• Your Sources (original data)
• Your Views (transformations created by the AI)
• Your Visualizations (charts and dashboards)
• Everything stays organized in one place

Create Your First Workbook

1. Click the "New Workbook" button

2. Give your Workbook a name (e.g., "My First Analysis")

3. Your fresh Workbook is ready!

---

Part 2: Import Data as a Source (3 min)

What are Sources?

Sources are your original, immutable data:

• CSV files, Excel spreadsheets, or database connections
• They live in the sources schema
• Shadowfax never modifies your Sources—they're read-only
• You'll see them in the "Sources" section of Workbook Objects

Import Your First Source

1. Click "Import Data" or drag-and-drop a CSV file

2. Upload your file and watch the progress

3. Preview your data to confirm it looks correct

4. Your Source appears in the sidebar under "Sources"

Schema Discovery

Shadowfax automatically analyzes your data:

• Detects data types and structures
• Identifies relationships between tables
• Builds a visual schema graph

---

Part 3: Your First Conversation (5 min)

Understanding @Mentions

To tell the AI which data to work with, use @mentions:

@[source_name]  or  @[view_name]

Type @ and you'll see an autocomplete dropdown with your Sources and Views.

Try Your First Prompt

Let's ask the AI to calculate monthly revenue. Type this in the Agent Chat:

@[orders] Calculate total revenue by month

What Happens Next

1. The agent thinks and plans

2. It creates a SQL View automatically

3. You see the result: A new View named something like monthly_revenue

Understanding Views

Views are SQL transformations created by the AI. Each View is:

• Reusable: Reference it in future prompts with @[view_name]
• Transparent: Click to see the SQL code and data
• Classified: Tagged by transformation type (Aggregate, Filter, Combine, etc.)

Click on your new View to inspect it:

Where's My View?

Find all your Views in the Workbook Objects sidebar under "Views":

Notice the "Aggregate" badge? That's the View's classification—it tells you what type of transformation it performs.

The Reactive System in Action

Views can depend on other Views or Sources. When you change something upstream, dependent Views automatically update.

Click the graph icon to see View dependencies:

Example:

• You create cleaned_orders (filters bad data)
• Then create monthly_revenue (aggregates from cleaned_orders)
• If you update cleaned_orders, monthly_revenue automatically recalculates

No broken formulas. No manual updates.

---

Part 4: Create Visualizations

Visualize Your Data

Use the /visualize command to create charts:

/visualize @[monthly_revenue] Show as a line chart with trend

Refine It

Keep iterating until it's perfect:

Add data labels and format Y-axis as currency

About Visualizations

• Powered by Vega-Lite grammar
• Unlimited flexibility—any chart type you can imagine
• Iterate conversationally until perfect
• Stored in the "Visualizations" section

---

Part 5: Helpful Slash Commands

Shadowfax has three powerful commands that change how the AI works:

/plan - Review Before Executing

When to use: Complex multi-step analyses where you want to approve the approach first.

/plan @[sales_data] @[customer_data]
Create a customer segmentation based on purchase behavior

What happens:

• AI shows you a detailed plan with steps
• You review and approve (or adjust)
• AI executes only after your confirmation

Why it's useful:

• Catch misunderstandings before execution
• Learn how the AI approaches problems
• Ensure alignment with your expectations

---

/visualize - Create Charts

When to use: Any time you want to create or modify a visualization.

/visualize @[product_performance]
Create a scatter plot with revenue on X, margin on Y,
bubble size = units sold, color by category

What happens:

• Activates specialized visualization mode
• Leverages advanced Vega capabilities
• Iterate until the chart is perfect

Why it's useful:

• Create sophisticated charts from natural language
• No manual chart builder wrestling
• Publication-ready visualizations

---

/recommendations - Get AI Suggestions

When to use: You're stuck or unsure what to analyze next.

/recommendations

I've just imported my sales data. What should I analyze first?

What happens:

• AI analyzes your current data and context
• Suggests relevant analyses or explorations
• Provides copy-paste prompts to try

Why it's useful:

• Discover insights you might not have considered
• Get unstuck when you don't know what's next
• Learn analytical techniques

/taxonomy - Discover Categories

When to use: You need to categorize or classify text data but aren't sure what categories exist.

/taxonomy @[support_tickets]

Classify tickets by customer intent based on subject and message body

What happens:

• AI explores your data and discovers natural patterns
• Proposes 3-7 meaningful categories with examples
• Shows you sample classifications for validation
• Creates an AI Table once you approve the taxonomy

Why it's useful:

• Automatically discovers categories without guessing
• Ensures you review classifications before they're applied
• Creates a reusable AI Table for consistent classification
• Perfect for organizing unstructured text data

---

Part 6: Understanding AI Tables

What are AI Tables?

AI Tables are special Views where AI adds intelligent columns using LLMs.

Use them for tasks SQL can't handle:

• Classification (e.g., categorize expenses, ticket priority)
• Extraction (e.g., extract entities from text)
• Sentiment analysis
• Summarization
• Any task requiring semantic understanding

Creating an AI Table

@[transactions]
Create an AI Table that categorizes each transaction
into expense categories: Travel, Food, Office, Software, Other

Important: AI columns start as NULL until you manually refresh the AI Table. Only you can trigger the AI processing—the agent can't do it automatically.

Example Use Case

Automatically categorize messy expense data:

Input columns: transaction_id, vendor_name, memo, amount AI output columns: category, subcategory, confidence_score

---

Part 7: Choose Your Adventure

Ready for a complete tutorial?

Pick a domain and follow a guided journey with sample data. Each tutorial takes 15-20 minutes and includes:

• Sample datasets (CSV files to download)
• 3 guided analysis paths
• Copy-paste prompts with expected results
• Real-world business insights

---

E-commerce & Sales Analytics

What you'll learn:

• Product performance analysis
• Customer segmentation (RFM)
• Sales trend forecasting

Sample data: Orders, Customers, Products

Start E-commerce Tutorial →

---

SaaS & Subscription Analytics

What you'll learn:

• MRR (Monthly Recurring Revenue) tracking
• Cohort retention analysis
• Churn risk scoring with AI Tables

Sample data: Users, Subscriptions, Usage Events

Start SaaS Tutorial →

---

Marketing & Campaign Analytics

What you'll learn:

• Campaign ROI measurement
• Funnel conversion analysis
• Multi-touch attribution modeling

Sample data: Campaigns, Leads, Conversions

Start Marketing Tutorial →

---

Finance & Operations Analytics

What you'll learn:

• Budget vs actual variance analysis
• Expense classification with AI Tables
• Cash flow forecasting

Sample data: Transactions, Budgets, Departments

Start Finance Tutorial →

---

Next Steps

Just want to explore on your own?

• Import your own data and start asking questions
• Reference the Prompt Engineering Guide for best practices
• Check the FAQ for common questions

Want to see what's possible?

• Browse Use Cases for inspiration
• Learn Tips & Tricks

---

Quick Reference: Key Concepts

Concept	What It Is	Why It Matters
Workbook	Your analytical workspace	Keeps all your work organized in one place
Sources	Your original data (CSV, DB connections)	Immutable—never modified by Shadowfax
Views	SQL transformations created by the AI	Reusable, transparent, automatically maintained
AI Tables	Views with AI-generated columns	Handle tasks SQL can't (classification, extraction)
@mentions	@[name] syntax in prompts	Tells AI which data to work with
Slash commands	/plan, /visualize, /recommendations, /taxonomy	Special modes that change AI behavior
Reactive system	Automatic dependency updates	Change upstream → downstream recalculates

---

Ready to build something? Pick an adventure above or start with your own data. Welcome to Shadowfax!